2024's Biggest Breakthroughs in Biology and Neuroscience

You have an ancestor that lived billions of years ago who goes by the name of LUCA. LUCA is the last universal common ancestor to all modern life on earth. Including us, including bacteria and frogs, fish, trees, fungi, everything is alive and it has a cell.

In a 2024 paper, researchers peered back in time to paint the most complete picture of LUCA yet, using evidence from a wide range of scientific disciplines. And come up with a holistic understanding of when LUCA existed and how it interacted with the environment. So what was LUCA like?

Ever since scientists realized that all life on Earth share certain characteristics that suggested to us that there was some common ancestor. Darwin. , n fact, in one of his letters it said that if you trace his idea of evolution back to the very beginning, that would imply that there was some common ancestor.

So it was really Darwin who first started thinking about what that universal ancestor could have looked like. LUCA was not the origin of life on Earth, but the emergence of life as we know it today. And Luca was not alone.

We can think of Luca as more like a population rather than like a single individual. And there probably were many different types of organisms living at that time, whose descendants later died out fairly shortly thereafter, or maybe died out only recently. But LUCA survived and went on to evolve into all modern life.

Today there are two domains of cellular life descended from Luca: prokaryotes and eukaryotes. Prokaryotes include bacteria and archaea, which are comparatively simple cells. Eukaryotes are more complex and include all forms of complex multicellular life.

We're paleontologists, most of us, we're interested in understanding early evolutionary history. However, we can only really get back at this ancestral point looking at modern organisms. To reveal our evolutionary history.

Edmund Moody, Phil Donahue and their team sought to reconstruct LUCA's genome or collection of genes. And then from these, try and build kind of like a working metabolic network and then use this as a basis to try and understand what LUCA would've been like. Their approach involved filling out an evolutionary family tree based on the genomic relationships among 700 living bacterial and archean species.

What we need to do is infer a tree of those 700 species, those 700 genomes. In order to do that, you need to pull out genes, which we think have evolved very slowly, been very conserved, haven't changed much over the billions of years. Using a suite of software and loads of computing power.

The team constructed a probabilistic gene tree leading back to LUCA. So this says, okay, well we've got X number of genes, a probability of 50% or a probability of 75%. Therefore, we think that these genes were probably in LUCA.

By integrating the probabilities of thousands of gene families, the team estimated LUCAs genome likely encoded 2,600 proteins. Our main take home result is that Luca is a complex organism, very similar, perhaps a bit smaller than a modern day bacterium, for example. LUCA would've had a simple phospholipid membrane and the molecular machinery for maintaining a genome and building proteins.

It could metabolize hydrogen gas and carbon dioxide. But one of the really interesting things that we found, which was a complete surprise to us, was this Crispr-Cas system. Crispr-Cas is a basic immune system cell's use to fight off viral attacks.

And some people are surprised that viruses would've been around that long ago. Luca may have lived off of non-living sources like hydrothermal vents or atmospheric gases. Or it may have dined on the chemical waste of other microbes.

To me, it seems more likely that LUCA would've been part of a complex ecosystem, even at that point in time, exchanging metabolic products, et cetera, with all sorts of different things. For a complete picture, the team also needed to determine when LUCA lived. And so we used some molecular and paleontological methods to try to estimate its age by calibrating the branches within genome trees.

By comparing the rates of mutation over time across gene families, they were able to arrive at a date. So we estimated Lucas's age to be around 4. 2 billion years old.

This is fairly soon after the earth would've cooled off and been habitable for life to emerge. To me, that really ancient age is very surprising because we reconstruct LUCA's being quite a complex organism, and yet it existed within about two, three, 400 million years of earth forming as a planet. Hopefully we've gone closer to the truth of this, but clearly more work will need to be done.

This is the vagus nerve. This long bundle of neurons is a two-way information superhighway that connects the brain to many of the body's internal organs. So there's a dizzying array of stimuli that comes up from the body and can evoke various behavioral and physiological responses.

The brain can also in turn regulate the body. This May, a team of scientists published research revealing a surprising new vagal connection between the brain and the body's immune system. And there is no question that the whole set of diseases and disorders that were previously thought to be diseases of the body are undoubtedly going to emerge as diseases of body brain communication.

In 1921, physiologist Otto Loewi discovered that stimulating a frog's vagus nerve slowed its heartbeat. He named the associated signaling chemical "vagusstoff", and later won the Nobel Prize. The vagus nerve contains some of the longest neurons in the body and connects the brainstem to vital organs including the stomach, lungs, and heart.

There are both sensory neurons that take information from the body up to the brain as well as motor neurons that go in the opposite direction. Through the vagus nerve, The brainstem regulates basic survival systems such as breathing, heartbeats and hunger. One term that's often used with the vagus is homeostasis, and so any deviation from that homeostatic setpoint, the body brain signals will help bring the body back to that equilibrium.

In 2020, Charles Zuker's lab stumbled upon a surprising brain body connection. Their studies of taste receptors revealed that in mice, the sensation of sweetness comes not only from the tongue but also from the stomach. Now this opened up this world body-brain signaling.

What other silly examples are over brain control of body biology in a way that's completely unexpected? And we thought the immune system poses an amazing example. At the time, there was no evidence that the brain communicated directly with the immune system.

However, experiments performed in the 1990s by neurosurgeon Kevin Tracey pointed to the possibility of a connection. So the crazy experiment at the time where he did a sort of a course electrical stimulation of the vagus nerve and he found that stimulating the vagus nerve dampened the immune system, and he called this the "anti-inflammatory reflex". Inflammation is the immune system's first line of defense.

After sustaining an injury or infection, pro-inflammatory molecules are released to help fight pathogens. Then the release of anti-inflammatory molecules keeps the inflammation response in check, so it doesn't damage the body's own tissues. A heightened inflammatory response has been linked to a host of diseases and disorders.

Multiple Sclerosis, type-one diabetes, lupus, and I can go on and on. And also some metabolic disease like obesity and diabetes, overactive inflammation, definitely a bad thing. In a series of experiments, the Zuuker team set out to find whether there's a mechanism in the brain that controls the inflammatory process.

We reasoned that there has to be some sort of a homeostatic controls. First, the researchers identified the groups of neurons in mice that are activated during immunological challenges. They wondered what would happen if they use genetic techniques to activate these neurons?

They can choose to excite or inhibit. They discovered that these neurons in the brainstem acted like a volume dial for inflammation. The researchers can manipulate them to turn the inflammatory response up or turn it back down.

Two lines, one carries pro-inflammatory, one carries anti, that was a huge surprise, that it's really telling the brain what's happening on both sides of this homeostat. This was the first time that control of the immune system was located in the brain. Further research of this inflammatory homeostat in humans could lead to new treatments for the many diseases linked to inflammation.

And if you could find ways to tap and control the activity either positively or negatively for each of these different neuron types, you can have profound effects. I think it opens up a new wind in the way we think about how perhaps we can help make a difference. These are microscopic molecular machines essential to life on earth.

They've evolved over millions of years to perform a vast array of vital functions. These are proteins. They're the molecules that do the work.

They interact with other molecules, they build other molecules, they take molecules down. All these are at the end chemical reactions, and to understand this chemistry of life, you need to understand the structure of these molecules. For more than half a century, biologists have sought to unravel the enigma of how proteins fold to function.

The effort to doing this is enormous. You can think of it as something like a hundred thousand dollars in expense, a couple years of a PhD student's time. Really an enormous investment to get even a single structure.

During a recent grand challenge, a team at DeepMind used artificial intelligence to solve a key part of the protein puzzle. They trained a neural network to read the one dimensional molecular sequence of a protein and predict its 3D structure. For many proteins, AlphaFold 2's accuracy is 99%.

DeepMind's breakthrough opened the door to a new era of biology. People have said that biology is becoming a computational science, and this is certainly true. The AI revolution, you start applying things to problems you did not solve experimentally before.

Including the engineering of human-designed proteins to fix some of the world's biggest problems. A protein's specific molecular function is a product of its three dimensional folded shape. They'll fold up into a really precise shape, and they do that every time and that shape carries out a biological function.

If we have the 3D structure, then we can really begin to understand how these molecules behave and function. This origami like shape is rendered by the sequence of its primary structural components: amino acids. All proteins are built from 20 different flavors of amino acids connected in chains called polypeptides.

When first assembled inside a cell, proteins are unfolded. Their amino acids strung together like beads on the necklace. These amino acids can be arranged in countless configurations to form different proteins.

The recipe for a given protein's specific polypeptide sequence is encoded within a cell's DNA. In 1969, biologists Cyrus Leventhal observed a paradox for any protein, even small ones, there's an astronomical number of possible folding configurations. Yet proteins reliably fold into their functional shapes in less than a second.

The mystery behind this process became known as the protein folding problem. Google DeepMind was founded in 2007 to advance a nascent form of AI called deep learning. After their AI system successfully mastered Go and several other games, DeepMind's founder Demis Hassabis sought new challenges.

We're very early days in development of AI, but there are still many unsolved problems. The DeepMind team entered what has been called the Olympics of protein, folding: the critical assessment of structure prediction challenge or CASP. CASP participants take amino acid sequences and then attempt to predict the protein's 3D structures using a computer algorithm.

Experimental chemist John Jumper led the team in the development of their protein structure prediction algorithm AlphaFold2. They trained a deep learning neural network with data that described more than a hundred thousand known proteins. Both their amino acid sequences and their folded 3D structures.

Plus evolutionary data about the proteins. Built some of our understanding of proteins into it, what's called inductive bias. In the language of machine learning, it learned extraordinarily rapidly from data.

The data is processed by a series of powerful neural networks called transformers. After cycling through the whole algorithm, AlphaFold reveals a structure prediction along with a score of how confident it is in its predictions of different parts of the protein structure. In 2020, the DeepMind team entered AlphaFold In CASP 14, The algorithm's structure predictions came out on top.

It was really a shock. You're looking at these things like, can it really be that good? What's going on here?

By July, 2022, DeepMind had released the structure predictions for 218 million proteins, nearly all of those known in the world. Some called the protein problem, essentially solved. For many biologists it took months or years to accept the results.

But at the University of Washington, David Baker had been developing software to solve the protein folding problem for three decades. At that point, we really sort of said, okay, well, how can we take these new concepts and ideas and AI generally and apply it to protein design. Which is the process of synthesizing new and novel proteins.

For making brand new proteins, there are no genes that encode them, so we have to make synthetic genes that encode these proteins. Baker's goal is to design new proteins that don't exist in nature to solve difficult problems facing humanity. The work we're doing sort of roughly falls into three general areas.

So the first is medicine, the second is energy and sustainability, and the third is sort of new technology. To design a new protein, researchers run the protein folding problem backwards. Instead of predicting the 3D shape of a protein from its amino acid sequence, they design the 3D shape for a novel protein and then use AI tools to output the amino acid sequence, before synthesizing the protein in the lab.

We can now design proteins which are much more sophisticated and should be much more precise and safe. Beyond medicine, we're working on improved methods of capturing sunlight and doing things with that energy. We're working on improved methods for degrading toxic compounds.

The next frontier in the application of AI to problems in protein science lies in the prediction of protein interactions within the whole cell. Right, these are the machines of the cell. They do lots of really, really important stuff in the cell.

Inside a cell, proteins interact with a host of different molecules, including DNA, RNA, and metals. So Baker's team, DeepMind and others started developing AI algorithms capable of predicting these complex interactions. The spring of 2024 saw the release of the next generation of AI prediction tools.

The Baker Lab released RoseTTAFold All-Atom, which predicts the 3D structures of assemblies, of proteins and other small molecules. Soon after DeepMind released AlphaFold3. Really incredible improvements that we think are going to unlock a lot of new science.

The efforts of these three researchers and their teams have revolutionized the study of proteins. In October, 2024, the three were jointly awarded the Nobel Prize in chemistry to recognize their scientific leap forward. You always want to be pushing the frontier.

At a time of technology transition, it's a very intense time. For me, It's been tremendously exhilarating. I mean, there's so much more to be understood.

It's the beginning.